yuriyvnv/whisper-large-v3-high-mixed-pt

Whisper-Large-v3 Portuguese - High-Quality Filtered Synthetic Data

This model is a fine-tuned version of openai/whisper-large-v3 for Portuguese automatic speech recognition (ASR). It was trained on Common Voice 17.0 Portuguese combined with WAVe-filtered high-quality synthetic speech data only using a strict threshold (q ≥ 0.8).

Purpose

This model demonstrates the effectiveness of quality-over-quantity filtering for synthetic speech data. By retaining only the top 33.3% of synthetic samples (those with WAVe scores ≥ 0.8), this model achieves:

• 32.6% WER improvement over the CV-only baseline (7.94% vs 11.78%)
• 18.9% better cross-domain generalization on MLS (12.41% vs 15.31%)
• 34% increase in training steps vs baseline (575 vs 430)

The model is part of a comprehensive study on WAVe (Word-Aligned Verification) filtering for Portuguese ASR, demonstrating that strict quality filtering provides an optimal balance between performance gains and computational efficiency.

Model Details

Property	Value
Base Model	openai/whisper-large-v3
Language	Portuguese (pt)
Task	Automatic Speech Recognition (transcribe)
Parameters	1550M
Training Data	Common Voice 17.0 + High-Quality Synthetic (q ≥ 0.8)
Total Training Samples	29,178
Sampling Rate	16kHz

Evaluation Results

This Model (whisper-large-v3-high-mixed-pt)

Metric	Value
Validation Loss	0.1045
Validation WER	7.33%
Test WER (Common Voice)	7.94%
Test WER (MLS)	12.41%
Best Checkpoint	Step 200
Max Training Steps	575

Comparison with Other Training Configurations (Whisper-Large-v3 Portuguese)

Training Data	Max Steps	Val Loss	Val WER	Test WER (CV)	Test WER (MLS)
Common Voice Only	430	0.1260	11.38%	11.78%	15.31%
High-Quality (q ≥ 0.8) + CV	575	0.1045	7.33%	7.94%	12.41%
Mid-High (q ≥ 0.5) + CV	805	0.1040	7.73%	8.33%	10.27%
All Synthetic + CV	860	0.1050	7.57%	8.33%	13.43%

Key Performance Highlights

• Best in-domain performance: Lowest Test WER (7.94%) on Common Voice among filtered models
• Strong cross-domain: 18.9% relative improvement on MLS vs baseline
• Most efficient filtering: Only 33.5% more samples than baseline, 33% fewer than unfiltered
• Optimal quality-to-compute ratio: Achieves near-best performance with minimal synthetic data

Training Data

Dataset Composition

Source	Samples	Description
Common Voice 17.0 Portuguese	21,866	Real speech from Mozilla's crowdsourced dataset
Synthetic Transcript PT (q ≥ 0.8)	7,312	Strictly WAVe-filtered TTS audio (high quality only)
Total	29,178

Synthetic Data Generation Pipeline

The synthetic dataset (yuriyvnv/synthetic_transcript_pt) was generated using:

1. Transcript Generation: GPT-4o-mini, matching Common Voice word count distribution
2. Speech Synthesis: OpenAI TTS-1 model with 9 voice variants (alloy, ash, coral, echo, fable, nova, onyx, sage, shimmer)
3. Quality Filtering: WAVe model with strict threshold q ≥ 0.8

WAVe Quality Distribution (Portuguese Synthetic Data)

Quality Level	Samples	Percentage	Used in This Model
High (q ≥ 0.8)	7,312	33.3%	✓
Medium (0.5 ≤ q < 0.8)	11,869	54.0%	✗
Low (q < 0.5)	2,787	12.7%	✗

This strict threshold retains only the top 33.3% of synthetic samples, prioritizing quality over quantity for maximum training efficiency.

Training Procedure

Hyperparameters

Parameter	Value
Learning Rate	5e-6
Batch Size (Global)	256
Warmup Steps	200
Max Epochs	5
Precision	BF16
Optimizer	AdamW (fused)
Eval Steps	50
Metric for Best Model	eval_loss

Training Infrastructure

• GPU: NVIDIA H200 (140GB VRAM)
• Operating System: Ubuntu 22.04
• Framework: Hugging Face Transformers

Usage

Transcription Pipeline

from transformers import pipeline

transcriber = pipeline(
    "automatic-speech-recognition",
    model="yuriyvnv/whisper-large-v3-high-mixed-pt",
    device="cuda"
)

result = transcriber("path/to/portuguese_audio.wav")
print(result["text"])

Direct Model Usage

from transformers import WhisperProcessor, WhisperForConditionalGeneration
import librosa

processor = WhisperProcessor.from_pretrained("yuriyvnv/whisper-large-v3-high-mixed-pt")
model = WhisperForConditionalGeneration.from_pretrained("yuriyvnv/whisper-large-v3-high-mixed-pt")
model.to("cuda")

audio, sr = librosa.load("path/to/portuguese_audio.wav", sr=16000)
input_features = processor(audio, sampling_rate=16000, return_tensors="pt").input_features.to("cuda")

predicted_ids = model.generate(input_features)
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
print(transcription)

Specifying Language

model.generation_config.language = "pt"
model.generation_config.task = "transcribe"

Methodology

This model leverages WAVe (Word-Aligned Verification), a word-level quality assessment method for filtering synthetic speech data. Unlike sentence-level filtering approaches, WAVe:

• Aligns each word to its corresponding audio frames using multi-head attention
• Assigns per-word confidence scores via a GLU-based scorer
• Detects localized synthesis errors (mispronunciations, omitted words, prosodic anomalies)
• Achieves 6.5% improvement over sentence-level filtering methods

The strict threshold (q ≥ 0.8) retains only the top 33.3% of synthetic samples, ensuring that only the highest-quality synthetic speech is used for training.

When to Use This Model

This model is ideal when:

• Best in-domain accuracy required: Achieves 7.94% WER on Common Voice Portuguese
• Compute efficiency matters: 33% fewer synthetic samples than unfiltered approach
• Quick fine-tuning needed: Smaller dataset (29,178 samples) enables faster iteration
• Quality over quantity: Only top-tier synthetic data (33.3%) for clean training signal

Consider other variants based on your needs:

• whisper-large-v3-mixed-pt: Better cross-domain performance (10.27% MLS)
• whisper-large-v3-cv-fully-synthetic-pt: Maximum data augmentation

Limitations

• Domain specificity: Optimized for general Portuguese; may underperform on technical domains
• Acoustic conditions: Trained on clean speech; noise robustness not guaranteed
• Dialect coverage: Performance may vary across Portuguese regional variants (European vs Brazilian)

Citation

This model is part of research on WAVe (Word-Aligned Verification) for synthetic speech quality assessment. While the WAVe methodology paper is currently under review, please cite our previous work that motivated this research:

@article{perezhohin2024enhancing,
  title={Enhancing Automatic Speech Recognition: Effects of Semantic Audio Filtering on Models Performance},
  author={Perezhohin, Yuriy and Santos, Tiago and Costa, Victor and Peres, Fernando and Castelli, Mauro},
  journal={IEEE Access},
  year={2024},
  publisher={IEEE}
}

References

• Base Model: openai/whisper-large-v3
• Training Data (Real): mozilla-foundation/common_voice_17_0
• Training Data (Synthetic): yuriyvnv/synthetic_transcript_pt
• Whisper Paper: Robust Speech Recognition via Large-Scale Weak Supervision
• Motivating Research: Enhancing ASR with Semantic Audio Filtering (IEEE Access 2024)

License

Apache 2.0